XML Dispatch Test

These tests exercise your browser to see how it handles application dispatch (using both XML namespaces and stylesheets) in generic XML content served as application/xml.

As mentioned, on my weblog, the application/xml spec is ambiguous regarding how namespace dispatch should occur, so there's no right answer (though there are implications to the choice).

I'd appreciate sending me your browsers results so that the upcoming revision to 3023 can reflect common practice.

See also; Juicy Studio MIME type tests, and Mimasa's XHTML/HTML media type tests.


Test 1; Gibberish XML with a gibberish namespace

Test 2; XML labelled with the XHTML namespace

Test 3; XML that looks like XHTML, but has no namespace declaration

This one is also ambiguous per 3023, but IMO, it should have stated that this should never be processed as HTML/XHTML.

Test 4; Gibberish XML (except for one HTML tag) in gibberish namespace plus XSLT stylesheet

Test 5; XHTML plus XSLT stylesheet

Test 6; Gibberish XML with CSS stylesheet

  • Mozilla Firefox 1.0 and 1.5 display the numbers in line item form
  • IE 6.0/WinXP displays the numbers in line item form
  • Observations

    IE's behaviour regarding namespace dispatch alone is consistent, yet when a no-op stylesheet is applied, either the media type appears to be implicitly coerced or namespace-dispatch is activated (or some other funkiness is occuring). While this isn't breaking any specs, it seems to me to be unintended behaviour on the part of the browser.

    Safari's behaviour in test 1 is different, though arguably reasonable and accurate per RFC 3023. However, in test 3 - which I didn't expect any browser to choke on - it rendered text as XHTML without there being any indication in the media type or namespace that this was to be processed as XHTML. This is incompatible with the recommendations made in the TAG finding on authoritative metadata.


    It seems that RFC 3023 remains quite accurate (though non-specific) about namespace and stylesheet based dispatch; neither can be assumed for an arbitrary agent on the Web, although if Microsoft brings IE7 in line with this behaviour, then it seems worth documenting, despite the architectural problems.


    Thanks to the following people for their help;